Episode 131      25 min 14 sec
What seems to be the antimatter?: Where experimental particle physics meets cloud computing

Physicist Associate Professor Martin Sevior and software engineer Tom Highfield explain how commercial cloud computing can be enlisted in the service of answering questions about the origins of the universe. With science host Shane Huntington.

"The advantage of using cloud is that we also find that our load CPU demand over a year isn't constant. There are peaks and there are troughs. If we priced our purchase to satisfy our peak needs, we'd find that our system would lay idle for some fraction of the year." -- Associate Professor Martin Sevior




           



Associate Professor Martin Sevior
Associate Professor Martin Sevior

Martin Sevior obtained his Ph.D. in the field of Nuclear Astro-Physics from the University of Melbourne in 1984. In 1985 he worked at the TRIUMF cyclotron accelerator in Vancouver, Canada. In 1993 he returned to the University of Melbourne and is now working in the field of Experimental Particle Physics. He performs experiments with the world's highest intensity and energy particle accelerators in Japan and at CERN in Switzerland. These experiments investigate the cause of the Universal Matter-AntiMatter asymmetry (at the Belle experiment in Japan) and the origin of mass at the ATLAS at the CERN  laboratory in Switzerland. Both experiments probe conditions that last existed less than 1 billionth of a second after the Big Bang.

He has published over 420 papers in refereed Journals and has supervised 12 Ph.D. students to completion.  Martin’s core technical expertise is in the field of software. As such he has led Australia’s contributions to computing grid for high energy physics research.

In addition Martin is heavily involved in the Free Software community and is a core developer of the AbiWord multi-platform word processor.

In the middle of 2005 Martin and his colleagues at the School of Physics decided that they could make a useful contribution to the Nuclear Energy debate in Australia. They launched a website, nuclearinfo.net, describing their findings for all the issues relating to Nuclear Power in December 2005.  The site has been under continuous review since then and is updated as the team learns more.

Finally Martin maintains a blog (http://msevior.livejournal.com)which he updates from time to time.

Tom Fifield
Tom Fifield

Tom Fifield is a software engineer, based at The University of Melbourne in Australia. After gaining experience in grid computing working to support ATLAS at the Large Hadron Collider, Tom is now working extensively with collaborators from numerous overseas locations to facilitate the Belle II experiment's distributed computing design, and investigating interoperability between grid and cloud based solutions.

Martin Sevior obtained his Ph.D. in the field of Nuclear Astro-Physics from the University of Melbourne in 1984. In 1985 he worked at the TRIUMF cyclotron accelerator in Vancouver, Canada. In 1993 he returned to the University of Melbourne and is now working in the field of Experimental Particle Physics. He performs experiments with the world's highest intensity and energy particle accelerators in Japan and at CERN in Switzerland. These experiments investigate the cause of the Universal Matter-AntiMatter asymmetry (at the Belle experiment in Japan) and the origin of mass at the ATLAS at the CERN  laboratory in Switzerland. Both experiments probe conditions that last existed less than 1 billionth of a second after the Big Bang.

He has published over 420 papers in refereed Journals and has supervised 12 Ph.D. students to completion.  Martin’s core technical expertise is in the field of software. As such he has led Australia’s contributions to computing grid for high energy physics research.

In addition Martin is heavily involved in the Free Software community and is a core developer of the AbiWord multi-platform word processor.

In the middle of 2005 Martin and his colleagues at the School of Physics decided that they could make a useful contribution to the Nuclear Energy debate in Australia. They launched a website, nuclearinfo.net, describing their findings for all the issues relating to Nuclear Power in December 2005.  The site has been under continuous review since then and is updated as the team learns more.

Finally Martin maintains a blog (http://msevior.livejournal.com)which he updates from time to time.

Credits

Host: Shane Huntington
Producers: Kelvin Param, Eric van Bemmel
Series Creators: Eric van Bemmel and Kelvin Param
Audio Engineer: Gavin Nebauer
Voiceover: Nerissa Hannink

View Tags  click a tag to find other episodes associated with it.

Download file Download mp3 (24.3 MB)

What seems to be the antimatter?: Where experimental particle physics meets cloud computing


VOICEOVER
Welcome to Up Close, the research, opinion and analysis podcast from the University of Melbourne, Australia.

SHANE HUNTINGTON
I'm Shane Huntington. Thanks for joining us. When studying the natural world, we often find processes and relationships of high complexity. When we represent them in mathematical or digital terms, we get extraordinary numbers of variables and very large data sets. One such area where the amount of data is truly staggering is experimental particle physics. While physicists naturally turn to super computers to help them with their work, there are limitations to processing power and the costs can often be prohibitive.
Instead, researchers are starting to explore new possibilities, such as grid computing and commercially available cloud computing. To tell us more about how particle physicists address the enormous data challenge, we are joined by Associate professor Martin Sevior and Mr Tom Fifield, both from the School of Physics, here at the University of Melbourne, Australia. Welcome to Up Close, Martin and Tom.

MARTIN SEVIOR
Thank you very much, Shane. It's a pleasure to be here.

TOM FIFIELD
Thanks Shane.

SHANE HUNTINGTON
Martin, why don't we start with you because we want to initiate this discussion around antimatter and matter and why we're interested in these. Briefly, can you tell us what is antimatter?

MARTIN SEVIOR
In answering your question, I have to talk a little bit about the structure of matter itself. We know that matter is made of atoms; atoms are made of electrons, protons and neutrons. Antimatter are: antielectrons, antiprotons and antineutrons. The characteristics of these are that they have exactly the same mass as the electrons, protons and neutrons but they have opposite charge. An electron has negative charge; the antiparticle to an electron has positive charge. An antiproton has negative charge. Interestingly, an antineutron also has zero charge.

SHANE HUNTINGTON
Do we see antimatter in nature, either here on Earth or elsewhere in the universe?

MARTIN SEVIOR
Yes, we do see it Shane but it rises under special circumstances. For example, antimatter is widely used in medicine, in an imaging technique called positron emission tomography. So, this is called PET and in positron emission tomography, some radioactive nuclei actually emit antielectrons in their radioactive decay. These antielectrons collide with electrons and then annihilate, emitting two gamma rays. These gamma rays tell us very precisely where that nucleus actually was within the body. We can use this - it's a very powerful diagnostic tool. In addition, antimatter is created from cosmic ray interactions in the upper atmosphere.

SHANE HUNTINGTON
I'm assuming that there are antimatter equivalents of almost all matter particles. What occurs when an antimatter particle interacts with its corresponding matter particle? You mentioned that some have opposite charge obviously but there are some that don't have charge.

MARTIN SEVIOR
Almost always, the situation is that these particles annihilate one another. Using Einstein's famous mass energy equation, E = mc2, the mass energy of these particles is converted into energy. In the case of positron emission tomography, this is actually exploited. We get these two gamma rays - high energy photons - which basically come from the energy of the antielectron and an electron that it happens to find in regular nature.

SHANE HUNTINGTON
When we look around our world and around our universe, we see a lot of matter. We don't see any antimatter, certainly not visually, as we walk around. Why is there this massive imbalance in the universe, towards matter?

MARTIN SEVIOR
This is a really good question Shane because if you look at the fundamental equations of physics, they're very balanced between matter and antimatter. The inverse process that I talked about before, where matter and antimatter annihilate to create energy, happens also in nature. If you have a large amount of energy, you can convert that into matter but when you do that, you always have to make a particle and an antiparticle exactly at the same time. You could imagine, if you made an electron just all by itself, then suddenly you would have an extra negative charge in the universe.
Physics equations just don't allow that, so when a gamma ray converts itself into matter, it always converts into an electron and an antielectron, called a positron. The fundamental equations of physics imply that matter and antimatter should be created in equal and opposite amounts. If you wind back to the early universe, to the Big Bang, when energy was essentially used to create the matter that we see around us, you would naturally expect that matter and antimatter should be created in equal amounts. Then, you would naturally expect that the universe would have the same amount of matter and antimatter but of course, we don't see that.

SHANE HUNTINGTON
If that was actually the case - if we had a similar amount of matter and antimatter - presumably, whenever they interacted, they would collide. After a certain age, would the universe not just be filled with radiation or photons?

MARTIN SEVIOR
Very good question again, Shane. If you actually run this through and without this imbalance between matter and antimatter, what we find in our universe now is that there's around a billion times more photons than there are particles, or mass. If there wasn't any imbalance between matter and antimatter - they would be the same amount, matter and antimatter would collide all over the place. Except for little pockets here and there, where there happen to be a local imbalance of matter, local imbalance of antimatter. This ratio, instead of being one in a billion, would be one in a billion, billion. So essentially, there wouldn't be enough matter in the universe to make life. There wouldn’t be enough matter in the universe for there to be galaxies; there'd just be perhaps the odd star or two. Maybe very, very rarely, our universe would be an entirely different place and couldn't support life as we know it.

SHANE HUNTINGTON
The physics of the situation is very interesting but given this limitation on the amount of antimatter that we find in the universe, how do we actually go about studying it?

MARTIN SEVIOR
We study it by creating it in a laboratory; it's actually not that hard, as I said before. We just need enough energy and we can make it. For example, our experiment in Belle that we'll talk about later, collides positrons with electrons. We make those positrons by just running a very high energy electron beam into a target - creates electrons and positrons. We collect the positrons and then, we collide them later in a large accelerator.

SHANE HUNTINGTON
Both of you are working on the Belle experiment, which is partly based on particle accelerators in Japan. Tell us a bit about what the goals of the Belle experiment are and what the expectation is in the coming years from that.

MARTIN SEVIOR
Back to the question you asked about matter and antimatter in the universe, so this is a big issue. To physicists, they want to know the answers, to why we have more matter than antimatter. A clue was discovered in 1964, when it was discovered that a neutral long-lived particle - we say long-lived but it was only 50 billionths of a second, which is long to a particle physicist - was found to decay more frequently to a state of matter, as opposed to antimatter.
Following that discovery, a brilliant physicist, Andrei Sakharov, from the then Soviet Union wrote a seminal paper where he showed that this matter, antimatter imbalance in the universe could be explained, given this asymmetry in the way that matter and antimatter actually decay. That means how they actually fall apart and go from a heavy particle to a lighter particle. This was taken up in 1973 by Kobayashi and Maskawa, who found that they could explain this matter, antimatter asymmetry if the number of quarks in the universe was at least six. Now, this was an extremely bold suggestion because at the time, only three quarks had been found.

SHANE HUNTINGTON
And quarks are the components of protons and neutrons?

MARTIN SEVIOR
Exactly Shane, yes. Very interestingly, in 1974, the year after this paper made this prediction, the charm quark was discovered. In 1977, the bottom quark was discovered. In 1996, the top quark was discovered. So, their original prediction came true, in terms of the number of quarks but there was one missing piece. They also said that the way that the quarks interact with one another had to be very special and that had to have something that's called a complex component in the mixing of how the quarks change from one form to another. That hadn't been found by around 1998, when the Belle experiment started. So that's what we focussed on. We looked at creating B mesons, which were expected to show this matter, antimatter asymmetry in a way that would demonstrate unequivocally, that this complex mixing happens between quarks as well.

SHANE HUNTINGTON
The B meson - is this the particular particle that gives you a specific insight into this question?

MARTIN SEVIOR
Absolutely Shane. A B meson consists of a light quark, like the sort that make up protons and neutrons and a heavy quark, called a B quark. So, this was the one that was discovered in 1977. It has a mass of about five times that of a proton. When we create them, we create them in equal and opposite amounts, so we need a total amount of energy equal to 10 times that of a proton, which happens to be 10.5 GeV. One GeV is about the mass of a proton.

SHANE HUNTINGTON
Martin, tell us the origins of the term Belle, as in, in the Belle experiments. What is this experiment named after?

MARTIN SEVIOR
The Belle experiments was named after the fact that this B quark I talked about before actually stands for beauty. Another name for it is the beauty quark. Beauty in French is Belle, spelt Belle and if you look at that, that is almost the same forwards and backwards. Except, we put a B on the front to demonstrate we've seen an asymmetry in the word, the same way we're looking for an asymmetry in nature, between matter and antimatter.

SHANE HUNTINGTON
The Belle experiment looks at many individual events in a given day. About how many of these events are you trying to capture?

MARTIN SEVIOR
About 20 million per day, Shane.

SHANE HUNTINGTON
You mentioned the B meson before, as the particle of choice that you're after. Is it a bit of a needle in a haystack problem to find this particle within the experiment, or are they produced on mass?

MARTIN SEVIOR
It's no problem to find B mesons; what is a problem is to find them decaying exactly the right way. We're only interested in a small fraction of all of the different ways the B meson can decay. That problem is a needle in a haystack though.

SHANE HUNTINGTON
Tom, let me ask you some questions now, as we move into the data element of the Belle experiment. About how much actual raw data is the Belle experiment producing in a day or a year?

TOM FIFIELD
Over the 10 year running periods, it produced around five petabytes of data. A petabyte being 1000 terabytes - 1000 gigabytes makes a terabyte.

SHANE HUNTINGTON
You're moving into what's called the Belle 2 phase of this experiment. With the Belle 2 experiment, how much data do you think will be coming out of this when it's running?

TOM FIFIELD
We're looking at around 50 times what happened in Belle. You could probably approximate this to around 50 million DVD movies worth of data.

SHANE HUNTINGTON
An extraordinary amount. When you take that data and try and examine it, I understand one way to do it is through grid computing. Can you tell us a bit about what grid computing is and how that process works?

TOM FIFIELD
The grid computing is really the workhorse of these large international scientific collaborations these days. So, you have this problem where you have a machine like a particle accelerator and a detector, which is producing enormous amounts of data but you've got physicists or scientists in the general case, all around the world, who want to access that data. There's all of these common things that you find like how do I control who has access to what data? How do I move it from point A to point B? How do I share these resources? Making the sharing of resources seamless to scientists is the aim of grid computing.

SHANE HUNTINGTON
How does grid computing compare to just a standard super computer that we would find in various places around the world?

TOM FIFIELD
With a standard super computer, you'd typically have one interface to it and then you've got many, many users. With grid computing, you've got many, many heterogeneous resources. So, you've got different hardware vendors, you've got different software running all of these computing clusters. Grid computing provides a common interface to enable you to link all of these heterogeneous clusters.

SHANE HUNTINGTON
You mentioned software; presumably, there must be some software overlay of the entire system. How do you go about putting that together and how does that work?

TOM FIFIELD
The software overlay, we refer to as a middleware because it goes in between the users and the underlying resources. Its job is to abstract all of the details away from both sides of the system. The middleware has been developed over the last decade, starting from the early 2000s. There's been a significant number of large, government-funded projects, such as the Enabling Grids for E-science project, funded by the EU, as well as things like the Large Hadron Collider data grid and similar projects in the United States, such as the Open Science Grid.

SHANE HUNTINGTON
Tom, tell us, how does the cost of this grid structure compare to one or many super computers to do similar jobs?

TOM FIFIELD
It might seem counterintuitive, given that we're investing so much effort into it but grid computing actually involves significantly more overhead than running one central cluster. However, we find that in large scientific collaborations, such as those in particle physics, where we've got physicists from 10s of different countries, it's not financially or politically feasible for all of these countries to send all of this money into one particular place and build one giant computer system. This is one of the reasons why the grid computing model has actually improved the amount of resources available to scientists in every country. They can access the resources of every country without moving money across borders.

SHANE HUNTINGTON
Martin, we're focussing very much here on these great new experiments coming up through the Belle program and so forth. What large problems have been solved to date, using the grid computing methodologies?

MARTIN SEVIOR
The classic example is the start up of a Large Hadron Collider, which commenced operation in 2010. Almost immediately after data was available, physicists were finding new things. ATLAS, CMS and ALICE are three different experiments that all operated the Large Hadron Collider. The classic example was right at the end of 2010, when the Large Hadron Collider switched from colliding protons to colliding heavy ions.
Almost immediately, in two weeks after the first collisions, ATLAS has discovered the quark-gluon plasma in a very significant way. This was a really major discovery for the ATLAS experiment but it wasn't just the ATLAS experiment who saw it. The CMS experiment and ALICE all had evidence for the quark-gluon plasma, within two weeks of these collisions actually taking place. That could only happen because physicists had immediate access to the data that was available through the worldwide grid.

SHANE HUNTINGTON
I'm Shane Huntington. And my guests today on Up Close are Associate Professor Martin Sevior and Mr Tom Fifield. We're talking about grappling with nature's big problems with the help of cloud computing here on Up Close, coming to you from the University of Melbourne, Australia.
You mentioned we're about to move from what has been termed the Belle 1 experiment to the Belle 2 experiment. Martin, can you give us some context of what's happened over the last decade and what's about to happen in the new experiment?

MARTIN SEVIOR
Absolutely, it's been a very exciting time for us. Belle had a competitor experiments that started at almost exactly the same time. It's been quite exciting to compete with them. They operated at the Stanford Linear Accelerator Collider and the name of their experiment was BaBar. Between the two of us, we discovered that Kobayashi and Maskawa were correct. Kobayashi and Maskawa won the Nobel Prize in 2008. The really interesting thing about all of this, in the context of the matter, antimatter asymmetry is that although they're right in describing what we see now, here on Earth, that number or their theory fails dramatically to explain the matter, antimatter asymmetry.
They failed by about a factor of 10 billion. This implies that this interesting new principle in nature - in this matter, antimatter asymmetry that we haven't found yet. We have hints, both from our own experiments at Belle and BaBar and also from the theoretical framework of particle physics that we can discover this if we could just collect 50 times more data. That's the aim of the Belle 2 experiment - to look for this new principle of nature that we suspect is present because of this large matter, antimatter asymmetry in the universe. That's why we're going to all the effort to be able to analyse this massive, massive data set.

SHANE HUNTINGTON
Tom, we've just discussed grid computing but my understanding is you guys are also starting to use what's called cloud computing. Can you give us a bit of a description of what cloud computing is and why people are so excited about it?

TOM FIFIELD
To talk about cloud computing, one of the things you first need to look at is this idea of virtualisation, which is the idea that within one physical computer, I can run what appears to be another computer inside it. This is actually not a new concept; it's been around since the 1960s, when IBM was playing with it. It's come to the fore again today because it allows us to have one really powerful machine and share it among other users. This is important for cloud computing because it means commercial entities like Amazon can go and buy a whole lot of these large machines. For each machine, they can share it with different users. Those users can run whatever operating system they'd like to, be it Macintosh or Windows or the various flavours of the Linux operating system.

SHANE HUNTINGTON
When you have a problem - a particle physics problem - how is that divided up into this system, where you may have a piece from Amazon, a piece from perhaps Google, or a piece from a variety of other commercial suppliers?

TOM FIFIELD
That's been a very important project for us recently because as we mentioned earlier, grid computing was all about this seamless integration of resources. So, we had to extend those ideas that we were used to, to use things like Amazon's Elastic Compute Cloud, which of course, didn't have those standard interfaces. We started working with a team of people based in Barcelona, who were working on the LHCB experiment, also at the Large Hadron Collider in CERN. They had made a software framework, which allowed them to sit atop at least two different grid computing standards, which didn't interoperate and allowed their physicists to use those resources. They made this software framework, which was modular enough that you could just clip in something, which would enable you to use cloud resources. That’s what we developed over 2009 to today.

SHANE HUNTINGTON
Gentlemen, how does this computing compare to SETI, the search for extraterrestrial intelligence? Has been using for many years, where they farm out small pieces of a problem to a larger number of compliant users, who will give up some of their CPU space, to solve those problems.

TOM FIFIELD
Absolutely, so this is one of the key things that we have in particle physics. As you mentioned before, we've got these collisions, which we refer to as events. We're able to split up our computing in many cases, based on these events, into work units like the SETI@home, with their BOINC software would be able to do. Unfortunately, our data sizes in particle physics and our software as well, are a little bit larger than what we'd want to wish on the average home desktop user, which is why we require our own dedicated resources. The principle's the same: we're splitting up one really large problem into many smaller chunks.

SHANE HUNTINGTON
Martin, let me ask you, when we think about the economics of this situation, how does cloud computing compare to the other sorts of computing we've spoken about, like grid and super computers?

MARTIN SEVIOR
One of the other distinctions between grid and super computers is that super computers in general are what are called very highly parallel. So, you need specialised architectures and specialised components, in order to use these. We don't need that in particle physics because we can split our jobs up, as Tom described. That makes grid cheaper than it would otherwise be. The advantage of using cloud is that we also find that our load CPU demand over a year isn't constant. There are peaks and there are troughs. If we priced our purchase to satisfy our peak needs, we'd find that our system would lay idle for some fraction of the year.
So, the advantage of Cloud is that we can just purchase the CPU that we need exactly when we need it, instead of having to buy resources to cover our whole period of need. The other advantage is that because of the tremendous scale of the commercial cloud providers, they're actually extremely competitive, in terms of how much CPU you can buy from them. Even if we bought all of our CPU from them, we would still find them competitive with a local cluster, simply because of their enormous scale.

SHANE HUNTINGTON
Martin, just look into the future for a moment: what things should we be expecting to see, adding cloud computing to the Belle experiment, over the coming years? When do you think we'll be seeing those answers coming out?

MARTIN SEVIOR
Belle 2 will start in 2014 and we also are continuing to analyse our current data set, as Tom said, completed in 2010. We might be lucky and discover something very interesting actually in our current data set, or we might have to wait till Belle 2 is fully up and running. This is science at the cutting edge; I can't predict exactly what we'll see. If I could, it wouldn't be science at the cutting edge.

SHANE HUNTINGTON
Associate Professor Martin Sevior and Tom Fifield, both from the School of Physics, the University of Melbourne, Australia. Thank you for both being our guests on Up Close today and giving us a better understanding of how we're going to solve nature's big problems, using cloud computing.

MARTIN SEVIOR
Thank you, Shane. It's been a pleasure.

TOM FIFIELD
Thanks for having us.

SHANE HUNTINGTON
Relevant links, a full transcript and more info on this episode can be found at our website at upclose.unimelb.edu.au. Up Close is a production of the University of Melbourne, Australia. This episode was recorded on 17 February 2011. You can also hear associate professor Martin Sevior in episode three of this program, where he discusses nuclear power. Our producers for this episode were Kelvin Param and Eric van Bemmel. Audio engineering by Gavin Nebauer. Background research by Christine Bailey. Up Close is created by Eric van Bemmel and Kelvin Param. I'm Shane Huntington. Until next time, goodbye.

VOICEOVER
You've been listening to Up Close. For more information visit upclose.unimelb.edu.au. Copyright 2011. The University of Melbourne.


show transcript | print transcript | download pdf